A computational method and software to implement the method have been developed to sift through vast quantities of digital flight data to alert human analysts to aircraft flights that are statistically atypical in ways that signify that safety may be adversely affected. On a typical day, there are tens of thousands of flights in the United States and several times that number throughout the world. Depending on the specific aircraft design, the volume of data collected by sensors and flight recorders can range from a few dozen to several thousand parameters per second during a flight. Whereas these data have long been utilized in investigating crashes, the present method is oriented toward helping to prevent crashes by enabling routine monitoring of flight operations to identify portions of flights that may be of interest with respect to safety issues.

Experience has taught that statistically atypical flights often pose safety issues. Conventional methods of finding anomalous flights in bodies of digital flight data require users to pre-define the operational patterns that constitute unwanted performances. Typically, for example, a program examines flight data for exceedences (e.g., instances of excessive speed, excessive acceleration, and other parameters outside normal ranges). In other words, a conventional flight-data-analysis computer program finds only the patterns it is told to seek in flight data, and is blind to newly emergent patterns that it has not been told to seek. The present method overcomes this deficiency in that it does not require any pre-specification of what to look for in bodies of flight data. The method is based partly on the principle that it is necessary, not only to look for exceedences, but to go beyond exceedences, looking for more subtle data patterns that often cannot be prescribed in advance.

The method involves a series of processing steps that convert the massive quantities of raw data, collected during routine flight operations, into useful information. The raw data are progressively reduced using both deterministic and statistical methods. Multivariate cluster analysis is performed to group flights by similarity with respect to flight signatures derived from parameter values. The process includes analysis of multiple selected flight parameters for a selected phase of a selected flight, and for determining when the selected phase of the selected flight is atypical in comparison with the corresponding phases of other, similar flights for which corresponding data are available. For each flight, there is computed an atypicality score based partly on results from the cluster analysis. The distribution of atypicality scores of all flights is used to identify flights for examination.

The data from each day's flights are processed during the night and summarized in a document, called the "Morning Report," that includes a list of the 20 percent of flights having the highest atypicality scores, ranked in order of descending atypicality score. For each flight, the report includes a plain-language description of what makes the flight atypical. With the help of software designed to be intuitively useable, an analyst works through this list of flights to the finest level of detail where necessary, examining the characteristics that made them atypical, assessing their operational significance, and determining the need for further action.

This work was done by Irving Statler, Thomas Chidester, and Michael Shafto of Ames Research Center; Thomas Ferryman, Brett Amidan, Paul Whitney, Amanda White, Alan Willse, Scott Cooley, Joseph Jay, Loren Rosenthal, Andrea Swickard, Derrick Bates, Chad Scherrer, and Bobbie-Jo Webb of Battelle Memorial Institute; Robert Lawrence of Safe Flight; Chris Mosbrucker, Gary Prothero, Adi Andrei, Tim Romanowski, Daniel Robin, and Jason Prothero of ProWorks Corporation; Robert Lynch of Flight Safety Consultants; and Michael Lowe of the U.S. Navy. For further information, access the Technical Support Package (TSP) free on-line at www.techbriefs.com/tsp under the Information Sciences category.

This invention has been patented by NASA (U.S. Patent No. 6,937,924). Inquiries concerning rights for the commercial use of this invention should be addressed to

the Ames Technology Partnerships Division at (650) 604-2954.

Refer to ARC-15041-1